import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn.preprocessing import StandardScaler, MinMaxScaler, OneHotEncoder
from sklearn.cluster import KMeans
from sklearn.metrics import silhouette_score
from sklearn.decomposition import PCA
from sklearn.pipeline import Pipeline
from sklearn.compose import ColumnTransformer
from kneed import KneeLocator
import plotly.graph_objects as go
import plotly.express as px
from plotly.subplots import make_subplots
from sklearn.metrics import silhouette_samples
from yellowbrick.cluster import KElbowVisualizer
import matplotlib.cm as cm
import plotly.io as pio
pio.renderers.default = "notebook"

# Load Data
file_path = "C:/Users/Jun/Downloads/customer_segmentation_clustered.csv"  
# file_path = "C:/Users/Jun/Downloads/customer_segmentation_data.csv" 
df = pd.read_csv(file_path)
print(df.head())
print(df.info())
print(df.describe())

   Customer_ID  Age  Gender  Annual_Income  Spending_Score Subscription_Type  \
0            1   58    Male          49180              44             Basic   
1            2   38    Male          90341              52          Standard   
2            3   58  Female          90341              26          Standard   
3            4   38    Male          49180              76             Basic   
4            5   38  Female          49180              63             Basic   

   Purchase_Frequency  Customer_Tenure     Channel Product_Category  
0                   5                6    In-Store       Automotive  
1                  16                3    In-Store         Clothing  
2                  14                5    In-Store       Automotive  
3                   7                4    In-Store      Electronics  
4                   7                6  Mobile App      Electronics  
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 2000 entries, 0 to 1999
Data columns (total 10 columns):
 #   Column              Non-Null Count  Dtype 
---  ------              --------------  ----- 
 0   Customer_ID         2000 non-null   int64 
 1   Age                 2000 non-null   int64 
 2   Gender              2000 non-null   object
 3   Annual_Income       2000 non-null   int64 
 4   Spending_Score      2000 non-null   int64 
 5   Subscription_Type   2000 non-null   object
 6   Purchase_Frequency  2000 non-null   int64 
 7   Customer_Tenure     2000 non-null   int64 
 8   Channel             2000 non-null   object
 9   Product_Category    2000 non-null   object
dtypes: int64(6), object(4)
memory usage: 156.4+ KB
None
       Customer_ID          Age  Annual_Income  Spending_Score  \
count  2000.000000  2000.000000    2000.000000     2000.000000   
mean   1000.500000    39.390000   85417.319000       57.812000   
std     577.494589    14.649751   28163.319583       24.823653   
min       1.000000    22.000000   49180.000000       10.000000   
25%     500.750000    22.000000   49180.000000       40.000000   
50%    1000.500000    38.000000   90341.000000       60.000000   
75%    1500.250000    58.000000  116601.000000       78.000000   
max    2000.000000    58.000000  116601.000000       99.000000   

       Purchase_Frequency  Customer_Tenure  
count         2000.000000      2000.000000  
mean            10.453000         4.658500  
std              5.933076         2.220212  
min              1.000000         1.000000  
25%              6.000000         3.000000  
50%             10.000000         4.000000  
75%             14.000000         6.000000  
max             29.000000         9.000000

# Preprocessing (Feature Scaling & One-Hot Encoding)
numeric_features = ['Age', 'Annual_Income', 'Spending_Score', 'Purchase_Frequency', 'Customer_Tenure']
categorical_features = ['Gender', 'Subscription_Type', 'Channel', 'Product_Category']

preprocessor = ColumnTransformer([
    ('num', StandardScaler(), numeric_features),
    ('cat', OneHotEncoder(drop='first'), categorical_features)
])

X_transformed = preprocessor.fit_transform(df)
print("Transformed Data Shape:", X_transformed.shape)

# Convert transformed data into DataFrame for visualization
encoded_columns = preprocessor.named_transformers_['cat'].get_feature_names_out(categorical_features)
scaled_columns = numeric_features
all_columns = list(scaled_columns) + list(encoded_columns)
transformed_df = pd.DataFrame(X_transformed, columns=all_columns)

# Display the transformed data
print("Transformed Data Preview:")
print(transformed_df.head())

Transformed Data Shape: (2000, 17)
Transformed Data Preview:
        Age  Annual_Income  Spending_Score  Purchase_Frequency  \
0  1.270646      -1.287007       -0.556544           -0.919315   
1 -0.094906       0.174870       -0.234190            0.935162   
2  1.270646       0.174870       -1.281840            0.597984   
3 -0.094906      -1.287007        0.732872           -0.582137   
4 -0.094906      -1.287007        0.209046           -0.582137   

   Customer_Tenure  Gender_Male  Subscription_Type_Premium  \
0         0.604373          1.0                        0.0   
1        -0.747187          1.0                        0.0   
2         0.153853          0.0                        0.0   
3        -0.296667          1.0                        0.0   
4         0.604373          0.0                        0.0   

   Subscription_Type_Standard  Channel_Mobile App  Channel_Online  \
0                         0.0                 0.0             0.0   
1                         1.0                 0.0             0.0   
2                         1.0                 0.0             0.0   
3                         0.0                 0.0             0.0   
4                         0.0                 1.0             0.0   

   Product_Category_Beauty  Product_Category_Books  Product_Category_Clothing  \
0                      0.0                     0.0                        0.0   
1                      0.0                     0.0                        1.0   
2                      0.0                     0.0                        0.0   
3                      0.0                     0.0                        0.0   
4                      0.0                     0.0                        0.0   

   Product_Category_Electronics  Product_Category_Home & Kitchen  \
0                           0.0                              0.0   
1                           0.0                              0.0   
2                           0.0                              0.0   
3                           1.0                              0.0   
4                           1.0                              0.0   

   Product_Category_Sports  Product_Category_Toys  
0                      0.0                    0.0  
1                      0.0                    0.0  
2                      0.0                    0.0  
3                      0.0                    0.0  
4                      0.0                    0.0

# Calculate the maximum number of clusters using the rule of thumb: square root rule
num_rows = df.shape[0]  # Get the total number of rows in the dataset
rt = int(round(np.sqrt(num_rows), 0))

# Determine Optimal Clusters (Elbow & Silhouette Method)
wcss = []
rt = rt+1
for i in range(1, rt):
    kmeans = KMeans(n_clusters=i, init='k-means++', random_state=42)
    kmeans.fit(transformed_df)
    wcss.append(kmeans.inertia_)

# Plot the Elbow Method graph
plt.figure(figsize=(10, 6))
plt.plot(range(1, rt), wcss, marker='o', linestyle='--')
plt.title('Elbow Method')
plt.xlabel('Number of Clusters')
plt.ylabel('WCSS')

# Use KneeLocator to find the elbow point
knee = KneeLocator(range(1, rt), wcss, curve='convex', direction='decreasing')
optimal_clusters_elbow = knee.elbow

# Highlight the elbow point on the plot
plt.axvline(x=optimal_clusters_elbow, color='red', linestyle='--', label=f'Optimal k = {optimal_clusters_elbow}')
plt.scatter(optimal_clusters_elbow, wcss[optimal_clusters_elbow - 1], color='red', s=100, zorder=5)  # Highlight the point

plt.show()

print(f"Optimal number of clusters (Elbow Method): {optimal_clusters_elbow}")

Optimal number of clusters (Elbow Method): 10

# Silhouette Score
silhouette_scores = []
for i in range(2, rt):
    kmeans = KMeans(n_clusters=i, init='k-means++', random_state=42)
    kmeans.fit(transformed_df)
    score = silhouette_score(transformed_df, kmeans.labels_)
    silhouette_scores.append(score)

# Silhouette Polts
optimal_clusters_silhouette = np.argmax(silhouette_scores) + 2
max_silhouette_score = max(silhouette_scores)

plt.figure(figsize=(10, 6))
plt.plot(range(2, rt), silhouette_scores, marker='o', linestyle='--')
plt.axvline(x=optimal_clusters_silhouette, color='red', linestyle='--', label=f'Best k = {optimal_clusters_silhouette}')
plt.scatter(optimal_clusters_silhouette, max_silhouette_score, color='red', s=100, zorder=5)
plt.title('Silhouette Score')
plt.xlabel('Number of Clusters')
plt.ylabel('Silhouette Score')
plt.show()
def plot_silhouette(n_clusters, X):
    fig, axes = plt.subplots(2, 5, figsize=(20, 10))
    axes = axes.flatten()
    
    for i, k in enumerate(range(2, 12)):
        kmeans = KMeans(n_clusters=k, random_state=42)
        cluster_labels = kmeans.fit_predict(X)
        silhouette_avg = silhouette_score(X, cluster_labels)
        sample_silhouette_values = silhouette_samples(X, cluster_labels)
        
        y_lower = 10
        for j in range(k):
            ith_cluster_silhouette_values = sample_silhouette_values[np.where(cluster_labels == j)]
            ith_cluster_silhouette_values.sort()
            size_cluster_j = ith_cluster_silhouette_values.shape[0]
            y_upper = y_lower + size_cluster_j
            color = cm.nipy_spectral(float(j) / k)
            axes[i].fill_betweenx(np.arange(y_lower, y_upper), 0, ith_cluster_silhouette_values, facecolor=color, edgecolor=color, alpha=0.7)
            y_lower = y_upper + 10  
        
        axes[i].set_title(f'Silhouette Plot for k={k}')
        axes[i].set_xlabel('Silhouette Coefficient Values')
        axes[i].set_ylabel('Cluster Label')
        axes[i].axvline(x=silhouette_avg, color='red', linestyle='--')
        axes[i].text(0.9, 0.02, f'Silhouette: {silhouette_avg:.2f}', transform=axes[i].transAxes, fontsize=10, verticalalignment='bottom', horizontalalignment='right')
    
    plt.tight_layout()
    plt.show()

print(f"Maximum Silhouette Score: {max_silhouette_score:.2f}")

if max_silhouette_score < 0.5:
    print("Silhouette Score is less than 0.5. Omitting Silhouette Score.")
    optimal_k = optimal_clusters_elbow
else:
    print(f"Optimal number of clusters (Silhouette Score): {optimal_clusters_silhouette}")
    optimal_k = max(optimal_clusters_elbow, optimal_clusters_silhouette)

plot_silhouette(optimal_k, transformed_df)

Maximum Silhouette Score: 0.28
Silhouette Score is less than 0.5. Omitting Silhouette Score.

print(f"Final Optimal Number of Clusters: {optimal_k}")

Final Optimal Number of Clusters: 10

# Apply KMeans Clustering with Optimal K
kmeans = KMeans(n_clusters=optimal_k, random_state=42, n_init=10)
df['Cluster'] = kmeans.fit_predict(X_transformed)+1

# PCA for Visualization
# Perform PCA with 3 components
pca = PCA(n_components=3)
X_pca = pca.fit_transform(X_transformed)
df['PCA1'] = X_pca[:, 0]
df['PCA2'] = X_pca[:, 1]
df['PCA3'] = X_pca[:, 2] 

# Create a 3D scatter plot with Plotly
fig = go.Figure()

# Define colors for clusters
colors = px.colors.qualitative.Plotly  # Use Plotly's default qualitative colors

# Add data points for each cluster separately
for cluster in sorted(df['Cluster'].unique()):
    cluster_data = df[df['Cluster'] == cluster]
    fig.add_trace(go.Scatter3d(
        x=cluster_data['PCA1'],  # PCA Component 1
        y=cluster_data['PCA2'],  # PCA Component 2
        z=cluster_data['PCA3'],  # PCA Component 3
        mode='markers',
        marker=dict(
            size=5,  # Marker size
            opacity=0.7  # Transparency
        ),
        name=f'Cluster {cluster}',  # Legend label
        hovertext=cluster_data.index,  # Tooltip: Show customer ID or index
        hovertemplate='Customer ID: %{hovertext}<br>PCA1: %{x}<br>PCA2: %{y}<br>PCA3: %{z}'  # Custom tooltip
    ))

# Set the title and layout details
fig.update_layout(
    title=dict(text='3D Visualization of Customer Clusters in PCA Space', x=0.5),  # Centered title
    scene=dict(
        xaxis=dict(title='PCA Component 1', backgroundcolor="#fcf0dc", gridcolor='white'),
        yaxis=dict(title='PCA Component 2', backgroundcolor="#fcf0dc", gridcolor='white'),
        zaxis=dict(title='PCA Component 3', backgroundcolor="#fcf0dc", gridcolor='white'),
    ),
    width=900,  # Plot width
    height=800  # Plot height
)

# Show the plot
fig.show()

# pca = PCA(n_components=2)
# X_pca = pca.fit_transform(X_transformed)
# df['PCA1'] = X_pca[:, 0]
# df['PCA2'] = X_pca[:, 1]

# # PCA Visualization
# plt.figure(figsize=(10, 6))
# sns.scatterplot(x=df['PCA1'], y=df['PCA2'], hue=df['Cluster'], palette='tab10', alpha=0.7)
# plt.xlabel('PCA Component 1')
# plt.ylabel('PCA Component 2')
# plt.title('Customer Segmentation - PCA Visualization')
# plt.legend(title='Cluster')
# plt.show()

# # Cluster Distribution with Data Labels
# plt.figure(figsize=(8, 8))
# ax = sns.countplot(x=df['Cluster'], palette='tab10', hue=df['Cluster'], legend=False)  # Add hue to avoid warning
# plt.xlabel('Cluster')
# plt.ylabel('Number of Customers')
# plt.title('Cluster Distribution')

# # Add data labels
# for p in ax.patches:
#     ax.annotate(
#         f'{int(p.get_height())}',  # Text to display (count of customers)
#         (p.get_x() + p.get_width() / 2., p.get_height()),  # Position of the label
#         ha='center',  # Horizontal alignment
#         va='center',  # Vertical alignment
#         xytext=(0, 10),  # Offset from the bar
#         textcoords='offset points'
#     )

# plt.show()

# Cluster Distribution with Data Labels (2D)
plt.figure(figsize=(8, 8))
ax = sns.countplot(x=df['Cluster'], palette='tab10', hue=df['Cluster'], legend=False)  # Add hue to avoid warning
plt.xlabel('Cluster')
plt.ylabel('Number of Customers')
plt.title('Cluster Distribution')

# Add data labels
for p in ax.patches:
    ax.annotate(
        f'{int(p.get_height())}',  # Text to display (count of customers)
        (p.get_x() + p.get_width() / 2., p.get_height()),  # Position of the label
        ha='center',  # Horizontal alignment
        va='center',  # Vertical alignment
        xytext=(0, 10),  # Offset from the bar
        textcoords='offset points'
    )

plt.show()

# Analyze Customer Behavior
cluster_analysis_numeric = df.groupby('Cluster')[numeric_features].mean()
cluster_analysis_categorical = df.groupby('Cluster')[categorical_features].agg(lambda x: x.value_counts().index[0])
# cluster_analysis_numeric
# Combine numerical and categorical analysis
cluster_analysis = pd.concat([cluster_analysis_numeric, cluster_analysis_categorical], axis=1)
cluster_analysis

# Normalize the numeric features for the radar chart
scaler = MinMaxScaler()
cluster_analysis_normalized = pd.DataFrame(
    scaler.fit_transform(cluster_analysis_numeric),
    columns=cluster_analysis_numeric.columns,
    index=cluster_analysis_numeric.index
)

# Create a 2x5 grid of subplots
fig = make_subplots(
    rows=5, cols=2,  # 5 rows x 2 columns
    specs=[[{'type': 'polar'}] * 2] * 5,  # Each subplot is a radar chart
    subplot_titles=[f'Cluster {i}' for i in cluster_analysis_normalized.index]  # Titles for each subplot
)

# Add a radar chart for each cluster
for i, cluster in enumerate(cluster_analysis_normalized.index):
    row = (i // 2) + 1  # Calculate row position (1 to 5)
    col = (i % 2) + 1   # Calculate column position (1 or 2)
    
    fig.add_trace(
        go.Scatterpolar(
            r=cluster_analysis_normalized.loc[cluster].values,  # Values for each feature
            theta=numeric_features,  # Feature names
            fill='toself',  # Fill the area under the line
            name=f'Cluster {cluster}',  # Legend label
            line_color=px.colors.qualitative.Plotly[i],
            fillcolor=px.colors.qualitative.Plotly[i],
            hovertemplate='%{theta}: %{r:.2f}'  # Tooltip with feature name and value
        ),
        row=row, col=col  # Position in the grid
    )

# Update layout for the radar charts
fig.update_layout(
    title='Radar Charts of Customer Behavior by Cluster',
    showlegend=False,  # Hide legend (optional)
    polar=dict(
        radialaxis=dict(
            visible=True,
            range=[0, 1]  # Normalized range
        )
    ),
    width=1000,  # Adjust width
    height=2500  # Adjust height
)

# Show the grid of radar charts
fig.show()

# Category Distribution per Cluster Visualization
clusters_sorted = sorted(df['Cluster'].unique())
for cluster in clusters_sorted:
    total_count = df[df['Cluster'] == cluster].shape[0]
    fig, axes = plt.subplots(2, 2, figsize=(14, 10))
    fig.suptitle(f'Category Distribution in Cluster {cluster} (Total: {total_count})', fontsize=16)
    
    for ax, category in zip(axes.flatten(), categorical_features):
        category_counts = df[df['Cluster'] == cluster][category].value_counts()
        colors = sns.color_palette('tab10', len(category_counts))
        unique_categories = df[categorical_features].melt()['value'].unique()
        color_map = {val: 'blue' if val == 'Male' else 'pink' if val == 'Female' else color for val, color in zip(unique_categories, sns.color_palette('tab10', len(unique_categories)))}
        colors = [color_map[val] for val in category_counts.index]
        
        bars = ax.bar(category_counts.index, category_counts.values, color=colors)
        ax.set_xlabel(category)
        ax.set_ylabel('Count')
        ax.set_title(f'{category} Distribution')
        ax.set_xticks(range(len(category_counts.index)))
        ax.set_xticklabels(category_counts.index, rotation=45, ha='right')
        
        # Add data labels
        for bar in bars:
            ax.text(bar.get_x() + bar.get_width()/2, bar.get_height(), f'{int(bar.get_height())}', 
                    ha='center', va='bottom', fontsize=10, fontweight='bold')
    
    plt.tight_layout(rect=[0, 0, 1, 0.95])
    plt.show()

# for category in categorical_features:
#     category_counts = df.groupby(['Cluster', category])['Product_Category'].count().unstack()
#     ax = category_counts.plot(kind='bar', figsize=(12, 6), stacked=True)
#     plt.xlabel('Cluster')
#     plt.ylabel(f'Count of {category}')
#     plt.title(f'{category} Distribution Across Clusters')
#     plt.legend(title=category)
    
#     # Add data labels
#     for container in ax.containers:
#         ax.bar_label(container, fmt='%d', label_type='edge', fontsize=10, padding=3)
    
#     plt.show()

Cluster	Customer Segment	Description	Marketing Strategies	Recommendations
1	Premium Online Shoppers - Clothing Enthusiasts	Young, high-income shoppers who frequently buy fashion online.	- Exclusive online promotions - Loyalty programs - Influencer campaigns - AI-powered styling suggestions	- Flash sales & exclusive drops - Free shipping & priority delivery - Fashion subscription boxes
2	In-Store Home & Kitchen Shoppers - Budget-Conscious	Older, budget-conscious customers who prefer in-store purchases.	- In-store discount days - Coupon programs - Email promotions - Retail events & workshops	- "Buy More, Save More" program - Free home delivery for bulk purchases - Cashback incentives
3	Young Online Shoppers - Clothing & Beauty Lovers	Young, trend-driven online shoppers focused on fashion and beauty.	- Social media advertising - Gamified shopping experience - Limited-edition drops - Live shopping events	- Personalized beauty boxes - Interactive online fitting room - Subscription-based beauty bundles
4	In-Store Electronics & Home & Kitchen Buyers - Basic Subscribers	Middle-aged, practical buyers looking for essential home & tech products.	- In-store demo areas - Cross-promotional offers - Targeted SMS/email campaigns - Trade-in discounts	- Buy-now, pay-later financing - Extended warranties - Exchange program for old gadgets
5	In-Store Clothing & Home Buyers - Standard Subscribers	Balanced in-store and online shoppers who buy clothing and home essentials.	- Seasonal discount events - Multi-category bundling - VIP shopping hours - Referral rewards	- Frequent shopper loyalty program - Personalized in-store shopping assistance - QR-based shopping lists
6	Tech-Savvy Budget Shoppers	Online-focused, price-sensitive consumers who prioritize tech & deals.	- Flash sales & limited-time offers - Affiliate & cashback programs - AI-powered shopping bots - Subscription-based discounts	- Student discounts - Price-match guarantees - Budget bundle deals for electronics & clothing
7	Home & Kitchen Enthusiasts	Practical buyers focusing on home improvement and household goods.	- Product demonstration events - Bundled home deals - In-store financing plans - Personalized recommendations	- Customized home decor packages - Extended warranties for appliances - Upgrade & Save program
8	Premium Household Buyers	High-income, premium shoppers with a focus on luxury home products.	- Concierge shopping services - Luxe subscription plans - Exclusive VIP sales events - Sustainability-focused marketing	- Priority customer support - Complimentary installation services - White-glove delivery service
9	Digital Fashion Shoppers	Style-conscious online shoppers influenced by trends and social media.	- Social commerce integration - Influencer & trend-based promotions - AR fitting rooms - AI-powered style assistants	- Flash sales tied to trending fashion topics - Customized styling app - Virtual personal shopper service
10	Premium Electronics Buyers	High-value tech buyers who invest in premium gadgets and electronics.	- Extended warranty & protection plans - Trade-in programs - Tech concierge services - Early access & pre-order deals	- Personalized after-sales support - VIP launch event invitations - Subscription-based device protection plans

Customer Segmentation for Personalized Marketing Strategies¶

Table of Contents¶

1. Background ¶

2. Problem Statement ¶

3. Variables ¶

4. Feature Scaling & Encoding ¶

4.1. Feature Scaling (Standardization)¶

4.2. Encoding Categorical Variables ¶

4.3. Transformed Data Output ¶

5. K-Means Clustering ¶

5.1. Rule of Thumb: Estimating K Using the Square Root Rule ¶

5.2. Elbow Method ¶

5.3. Silhouette Score ¶

5.4. Silhouette Plots ¶

5.4.1. K=2 ¶

5.4.2. K=3 to K=5 ¶

5.4.3. K=6 to K=10 ¶

5.4.4. K=11 ¶

5.5. Optimal K ¶

6. Cluster Analysis ¶

6.1. 3D Visualization of Customer Segmentation Using PCA ¶

6.2. Cluster Distributiom ¶

6.3. Customer Behaviour Analysis ¶

6.3.1. Cluster 1: Premium Online Shoppers - Clothing Enthusiasts ¶

6.3.2. Cluster 2: In-Store Home & Kitchen Shoppers - Budget-Conscious ¶

6.3.3. Cluster 3: Young Online Shoppers - Clothing & Beauty Lovers ¶

6.3.4. Cluster 4: In-Store Electronics & Home & Kitchen Buyers - Basic Subscribers ¶

6.3.5. Cluster 5: In-Store Clothing & Home Buyers - Standard Subscribers ¶

6.3.6. Cluster 6: Tech-Savvy Budget Shoppers ¶

6.3.7. Cluster 7: Home & Kitchen Enthusiasts ¶

6.3.8. Cluster 8: Premium Household Buyers ¶

6.3.9. Cluster 9: Digital Fashion Shoppers ¶

6.3.10. Cluster 10: Premium Electronics Buyers ¶

7. Marketing Strategies & Recommendation ¶

8. Conclusion ¶

	Age	Annual_Income	Spending_Score	Purchase_Frequency	Customer_Tenure	Gender	Subscription_Type	Channel	Product_Category
Cluster
1	22.000000	113924.821656	84.980892	21.611465	2.407643	Female	Premium	Online	Clothing
2	58.000000	98280.069767	29.267442	7.647287	7.003876	Female	Standard	In-Store	Home & Kitchen
3	22.062992	99645.724409	84.141732	13.346457	2.397638	Male	Standard	Online	Clothing
4	37.929515	49180.000000	60.048458	8.343612	4.559471	Female	Basic	In-Store	Electronics
5	38.000000	96615.277311	60.033613	14.365546	4.436975	Male	Standard	In-Store	Clothing
6	22.579710	55227.673913	83.971014	4.775362	2.615942	Female	Basic	Online	Electronics
7	58.000000	49180.000000	29.222222	7.314815	7.101852	Male	Basic	In-Store	Home & Kitchen
8	58.000000	113961.105820	29.931217	7.259259	6.962963	Female	Premium	In-Store	Home & Kitchen
9	23.491525	49180.000000	83.152542	15.567797	2.542373	Male	Standard	Online	Clothing
10	37.687805	109939.926829	58.882927	6.775610	4.492683	Male	Premium	Online	Electronics